Graduate Thesis Or Dissertation


The Association Between Patient Activation and Cost, Utilization, and Outcomes Among Patients Undergoing Elective, Primary Total Knee Arthroplasty Public Deposited

Downloadable Content

Download PDF


Attribute NameValues
  • Background. Total knee arthroplasty, or replacement, is a common, generally successful, and expensive procedure. Tools to predict outcomes following orthopedic procedures are abundant, yet no commonly used assessment accounts for an individual’s propensity to engage in adaptive health behavior. The 13-item Patient Activation Measure (PAM) questionnaire is a tool that quantifies a patient’s ability and confidence in managing their healthcare needs. A patient who is highly activated has the knowledge and skill to navigate the healthcare system and engage in healthy behaviors. A patient who scores within PAM Level 4 has an ability to maintain healthy behaviors (highest level); Level 3 patients are taking action toward managing their health; Level 2 patients are building confidence; and Level 1 patients are not yet taking a role. The PAM assessment may help fill a pocket-of-need in the surgeon’s pre-operative evaluation of surgical readiness and post-acute risk. This study reports results of retrospective analysis of a double-blinded, prospective cohort study designed to assess the role of patient activation in a population of orthopedic surgical patients. The hypothesis is lower pre-operative patient activation scores are associated with higher costs incurred, higher post-operative service utilization, and lower patient-reported outcome measures in the first three months following total knee arthroplasty (TKA), commonly referred to as total knee replacement.   Methods. The study population includes 754 patients who elected to undergo primary, unilateral total knee arthroplasty, with no contralateral procedure during the study period, performed by surgeons at a single orthopedic clinic in Oregon, at two participating hospital sites between September 2014 and December 2015. Patients consented to be included in the research study under an alternative consenting process via an iPad application. Retrospectively, two different data sources were available: (1) patient survey data and (2) administrative billing data. Survey data were collected prospectively from patients pre-operatively and at three months (± one month) post-operatively. Pre-operatively, patients completed the Patient Activation Measure (PAM), an extremity-specific patient reported outcome instrument (Oxford Knee Score; OKS), a health-related quality of life tool (Veteran’s RAND VR-12 Health Survey; physical health component score and mental health component score, PCS and MCS, respectively), and pain assessment (Visual Analogue Pain Scale; VAS). A single de-identified data set with a unique study identification number for each patient, containing all billing data for the date of surgery through 90-days post-operatively, was received from the orthopedic clinic, one ambulatory surgery center, one regional hospital system (four hospitals), and a single community-hospital. Utilization counts, and payments deflated to 100% of Medicare, were summed for each patient for inpatient, outpatient, emergency department, and orthopedic clinic settings, and for a subset of patients participating in post-acute therapy services at the orthopedic clinic. Patient activation level (lowest - Level 1 & 2, Level 3, to highest - Level 4) and score (0-100, low to high), separately, were utilized as the explanatory variable of interest, with a theoretically-driven, identical set of covariates across all multivariate models. First, logit models were constructed to identify the probability of accessing services after hospital discharge (post-acute) across each care setting, including assessment of any hospital-based access. Models were constructed utilizing robust standard error and average marginal effects (AME), calculated utilizing bootstrapped standard errors to produce 95% confidence intervals, are reported. Next, ordinary-least-squares (OLS) was employed to estimate parameters in a series of regression models assessing: (1) total payments for the TKA episode of care (inpatient through 90-days post-operatively); (2) post-acute hospital payments (inpatient, outpatient, emergency department); (3) total post-acute payments, hospital plus the orthopedic clinic, excluding routine post-operative appointments; and (4) payments for therapy services. Duan’s smearing algorithm was applied to obtain a more accurate prediction of post-acute payments. Quantile-regression was used to further examine the effect of patient activation at different levels of payment for the total TKA episode of care (75th and 90th percentiles) and total post-acute payments (25th, 50th, 75th and 90th percentiles), utilizing bootstrapped standard errors. Due to limitations in the distribution of post-acute visits in the study population, only physical therapy services were examined in negative binomial regression models, where AME for PAM Levels is interpreted utilizing the finite-difference method. Finally, OLS was additionally utilized to test the hypothesis that lower patient activation is associated with lower scores on patient-reported outcome measures in the first three months after surgery. Overall 90-day post-operative score and change in score from pre- to post-operative assessment were examined. Approach to model building and reporting is consistent with payment models described above. Across all study aims and models, model specification and sensitivity tests were performed. Data were analyzed in Stata, version 14 (College Station, TX). This study was approved by the Oregon State University Institutional Review Board, Corvallis, OR, and PeaceHealth System Institutional Review Board, Eugene, OR. Results. The mean PAM score for the study population was 64.4 (sd = 12.8), with a median score of 60.6. Only 3% of patients fell within PAM Level 1 pre-operatively; 16% in Level 2; and the majority of patients had a pre-operative score within PAM Level 3 (55%) and Level 4 (26%). Overall, 23% of patients utilized at least one post-acute care service (n=174); 34 patients (4.5%) were readmitted; 110 accessed outpatient hospital or surgery center services (14.6%); and 65 patients (8.6%) had at least one ED visit. In multivariate models, neither patient activation functional form was not associated with increased probability of accessing post-acute hospital services when assessed separately. However, patients within PAM Level 3 have 8.1% higher probability of accessing any hospital-based post-acute care location compared to Level 4 patients (95% CI = 1%, 16%). Similarly, after adjusting for model covariates, patients in Level 3 have, on average, 2% higher total expenditures for the TKA episode of care than patients at the highest level of activation (95% CI = <1%, 4%). Patients in Level 3 also had 90% higher total post-acute payments (hospital plus clinic) compared to patients in the highest level of activation (95% CI = 11%, 227%). Patients at the highest level of activation (Level 4) have predicted total post-acute payments averaging $290 (sd = $402), compared to Level 3 patients who average a predicted $860 (sd = $407). Patients at the lowest two levels of activation have predicted total post-acute payments of $615 (sd = $248). Within the 75th percentile, patients who scored within PAM Level 3 have, on average, 167% higher post-acute payments compared to patients at the highest level of activation (95% CI = 0.3%, 610%). The effect is larger among the most expensive patients, those in the 90th percentile; Level 3 patients have 186% higher total post-acute payments (95% CI = 3%, 700%). Generally, patients in PAM Level 1 & 2 have positive coefficients in assessed models (i.e., higher than Level 4), but results were not statistically significant. There was no association between patient activation level or score and the initial length of hospital stay, after adjusting for all other potential confounders. Results of negative binomial regression demonstrated PAM Level 3 is significantly associated with the count of physical therapy visits (p=0.020). Compared to patients at the highest level of activation (Level 4), Level 3 patients have, on average, 1.79 more physical therapy visits (95% CI = 0.28, 3.28). Although not statistically significant, there is a gradient across the PAM Levels, with the most visits occurring among PAM Level 1 & 2 patients (1.87, 95% CI = -0.63, 4.38). There were 368 patients (48.8%) of patients who completed evaluable pre- and post-operative VR-12 questionnaires. Fewer patients (n=350, 46.4%) had paired pre- and post-VAS responses and the least number of patients completed both Oxford Knee Score (OKS) assessments (n=320, 42.4%). On average, patients in PAM Level 3 pre-operatively show 2.63 fewer points of improvement over time compared to Level 4 patients (95% CI = -5.08, -0.19). Patients in the two lowest levels of PAM have less improvement; 3.74 fewer than PAM Level 4 patients on average (95% CI = -7.19, -0.30). In contrast, for PAM Level as the predictor of interest, patients at lower activation levels demonstrate a greater improvement in MCS between pre- and post-operative assessments compared to Level 4 patients, but Level 4 patients had significantly higher MCS at baseline. There was no association between patient activation level, or score, and change in OKS over time. Among model covariates, pre-operative BMI and the performing surgeon resulted were significant predictors of OKS change. The amount of improvement, measured in the magnitude of score change, increases as BMI increases in the study population. For every 5-point increase in BMI, patients improve by two OKS points, on average (1.9 points, 95% CI = 1.3, 2.6). Results are consistent when assessing reductions in pain. For example, a patient with a pre-operative BMI equal to 25 would be expected to demonstrate a 2.4 point reduction in knee pain; BMI of 30 reports a 2.7 point reduction; and BMI equal to 35 reports 3.2 point reduction in pain between assessments. Both PAM Level and PAM score models yielded significant associations with VAS score. At the mean, the predicted change in pain assessment among Level 1 & 2 patients is -2.34 (sd = 0.86), or approximately a two-point reduction in pain post-operatively compared to pre-operative assessments. In contrast, Level 4 patients report, at the predicted mean, a 3.41 (sd = 0.77) reduction in self-reported pain. For every 10-point increase in PAM score the difference in VAS score between pre- and post-operative assessments grows increasingly more negative; 0.40 additional point reduction over time (95% CI = -0.59, -0.12). Across all patient-reported outcome measure, the performing surgeon was significantly associated with the magnitude of change between pre- and post-operative assessments. Discussion. This study is the first known investigation into the association between patient activation and post-acute payment and utilization following TKA. By combining electronic health records with administrative billing data, the current study successfully captured the majority of factors known to influence total knee arthroplasty outcomes. Under value-based care initiatives, providers can expect downward payment adjustments for complications resulting in post-acute utilization, and as a result of patients reporting lower gains on health-related quality of life metrics. Utilizing models predicting more than 1.3 million primary TKAs in 2020, PAM Level 3 patients will incur $784 million in additional total post-acute costs than patients who score within PAM Level 4 before surgery. This study is also consistent with prior research reporting higher patient activation leads to improved patient-reported physical health status and reductions in post-operative pain scores. While targeting medically complex patients for post-surgical optimization is an evidence-based strategy for managing post-acute costs, the current study demonstrates patient activation provides an additional piece of information useful for triaging patients into risk-based care coordination pathways. There remains only one other published study to assess patient activation and total joint arthroplasty functional outcomes, and this study is more representative of joint arthroplasty patients nationally. Although the distribution of patients within each PAM Level appears consistent with established research, the current pre-operative joint replacement education program may have systematically increased activation prior to baseline measurement. In addition, there is no way ascertain whether post-acute utilization was planned, or due to post-discharge complications from TKA. Data are limited to patients who were readmitted to the inpatient setting, or accessed another hospital-based service. Future work should seek to better understand and capture the relationship between patient activation and the population who were likely the most expensive post-operatively – those who were admitted to skilled nursing or rehabilitation centers following initial discharged to home. Conclusion. The orthopedic literature remains generally focused on medical and surgical predictors of “success” or “failure” following joint replacement surgery. Yet, after adjusting for many of the known risk factors for post-surgical complications, a patient’s level of activation – a measurement and patient characteristic unaddressed by the majority of orthopedic surgeons – explains differences in post-operative payments, utilization, and patient-reported health status. Two incentives are driving provider and delivery system change around total knee replacement; federal mandates to participate in bundled care payment programs for lower extremity reconstruction and the opportunity to shift select patients to outpatient ambulatory surgery centers. Both situations are propelling surgeons to adopt a population health framework for managing the TKA episode of care. Not all patients require the same level of support to achieve quality outcomes. The results of this study provide sound rationale for continuing to explore how patient activation influences patient experiences and outcomes for major elective surgical procedures, including TKA and more broadly total joint replacement.
Resource Type
Date Available
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Committee Member
Academic Affiliation
Non-Academic Affiliation
Rights Statement
Peer Reviewed
Additional Information
  • description.provenance : Approved for entry into archive by Steven Van Tuyl( on 2017-06-20T19:36:21Z (GMT) No. of bitstreams: 1OwenErinC2017.pdf: 3242837 bytes, checksum: dc68228cd2091a0483ff0dbf90062a29 (MD5)
  • description.provenance : Submitted by Erin Owen ( on 2017-06-16T17:43:06ZNo. of bitstreams: 1OwenErinC2017.pdf: 3242837 bytes, checksum: dc68228cd2091a0483ff0dbf90062a29 (MD5)
  • description.provenance : Made available in DSpace on 2017-06-20T19:36:22Z (GMT). No. of bitstreams: 1OwenErinC2017.pdf: 3242837 bytes, checksum: dc68228cd2091a0483ff0dbf90062a29 (MD5) Previous issue date: 2017-06-07
  • description.provenance : Approved for entry into archive by Julie Kurtz( on 2017-06-20T15:06:23Z (GMT) No. of bitstreams: 1OwenErinC2017.pdf: 3242837 bytes, checksum: dc68228cd2091a0483ff0dbf90062a29 (MD5)
Embargo reason
  • Existing Confidentiality Agreement
Embargo date range
  • 2017-10-21 to 2019-06-21



This work has no parents.

In Collection: